C-PRIME

Welcome

The vision of C-PRIME is to revolutionize predictive science by unraveling and leveraging the critical influence of small scales on emergent behaviors across various physical regimes. Our target application is a rotating detonation combustor (RDC)  which is a multiphysics, multiscale problem with a rich set of science challenges. We aim to transform how complex applications, from non-equilibrium multimaterial interface evolution to turbulence, are investigated by developing sophisticated autonomous artificial intelligence (AI) agents leveraging new inner- and outer-loop algorithms. These AI-driven frameworks will systematically refine the combinations of physics, simulations, algorithms, and computing infrastructure, significantly reducing the time from initial scientific inquiries to actionable predictive insights. 

Our main objectives are to:

• Extend the range of time and length scales accessible: Use multi-faceted adaptivity (solvers, methods, models, accuracy, precision), exploit data availability to develop machine learning (ML) based models, and use manifold theory, geometric integration, and new projection methods to accelerate solvers in time. 

• Advance probabilistic modeling of model-form errors: To build predictive multi-fidelity hierarchy, formally characterize modeling errors, use specially designed experiments to calibrate uncertainties, and invoke experimental design loops to optimally learn from data.

• Adapt solvers to hardware, and develop new hardware pathways for emerging solver frameworks: Represent solver interaction with hardware using digital twins to learn constraints, adapt algorithms to leverage hardware resources, and provide pathways for hardware innovation through non-Von Neumann architectures.

• Change the approach to scientific computing through an AI-enabled framework: Synthesize workflows that use solvers, algorithms, and data by using AI models to create specifications. Develop AI models, agents and synthesis framework. 

• Introduce cross-cutting verification: Besides validation and solution verification, build formal verification as a tool for human or AI-written codes, current/emerging hardware and hardware-software interfaces.

Funded by the Department of Energy under the Predictive Science Academic Alliance Program (PSAAP IV)